Skip to content

add fma #106

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 5 commits into
base: master
Choose a base branch
from
+6 −2
Open

add fma #106

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
f9ee8b0
add fma
JeffreySarnoff Jun 19, 2025
07fb447
test fma
JeffreySarnoff Jun 19, 2025
7493d9f
fix fma test
JeffreySarnoff Jun 20, 2025
7866260
Update Project.toml
JeffreySarnoff Jun 20, 2025
faad6d0
define fma using llvmcall
JeffreySarnoff Jun 20, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
name = "BFloat16s"
uuid = "ab4f0b2a-ad5b-11e8-123f-65d77653426b"
authors = ["Keno Fischer <[email protected]>"]
version = "0.5.1"
version = "0.5.2"

[deps]
LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
Expand Down
4 changes: 3 additions & 1 deletion src/bfloat16.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@ import Base: isfinite, isnan, precision, iszero, eps,
exponent_one, exponent_half, leading_zeros,
signbit, exponent, significand, frexp, ldexp,
round, Int16, Int32, Int64,
+, -, *, /, ^, ==, <, <=, inv,
+, -, *, /, ^, fma, ==, <, <=, inv,
abs, abs2, uabs, sqrt, cbrt,
exp, exp2, exp10, expm1,
log, log2, log10, log1p,
Expand Down Expand Up @@ -440,5 +440,7 @@ for F in (:abs, :abs2, :sqrt, :cbrt,
end
end

Base.fma(x::BFloat16, y::BFloat16, z::BFloat16) = BFloat16(fma(Float32(x), Float32(y), Float32(z)))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks very inefficient on hardware with native bf16 support.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Something like

ccall("llvm.fma.bf16", llvmcall, BFloat16, (BFloat16, BFloat16, BFloat16), x, y, z)

would be better (although this should probably be handled by julia's codegen directly)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why not just use llvmcall?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That would be better --
I was not sure it worked correctly (at one time it did not).
I am making the change.


# irrationals
BFloat16(x::AbstractIrrational) = BFloat16(Float32(x)::Float32)
2 changes: 2 additions & 0 deletions test/mathfuncs.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,8 @@ for F in (:abs, :abs2, :sqrt, :cbrt,
end
end

@test fma(phi, invphi, invphi) == BFloat16(fma(Float32(phi), Float32(invphi), Float32(invphi)))

for F in (:asec, :acsc, :cosh, :acosh, :acoth)
@eval begin
@test $F(phi) == BFloat16($F(Float32(phi)))
Expand Down
Loading